misr: a prototype new product -...

MISR: A Prototype New Product

Michael J. Garay

Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, USA

© 2014. All rights reserved.

International Cooperative for Aerosol Prediction (ICAP) 23 October 2014 Boulder, Colorado

Nine view angles at Earth surface: 70.5º forward to 70.5º backward Nine 14-bit pushbroom cameras 275 m - 1.1 km sampling Four spectral bands at each angle: 446, 558, 672, 866 nm 400-km swath: 9-day coverage at equator, 2-day at poles 7 minutes to observe each scene at all nine angles

2

Menu

3 3

• Appetizer – About the current MISR product

• Salad – Moving to 4.4 km resolution – NRT?

• Main – The MISR retrieval algorithm

• Dessert – Benford’s Law

Appetizer

4 4

Kahn et al. (2010). Multiangle Imaging SpectroRadiometer global aerosol product assessment by comparison with the Aerosol Robotic Network, Journal of Geophysical Research – Atmospheres

Africa

South America

United States Aug 29 Aug 26 Aug 25 Aug 24 Aug 23 Aug 22 Aug 21 Aug 20 Aug 19

Dust from the Sahara Desert Reaches Houston, Texas

Observations from the Multi-angle Imaging SpectroRadiometer (MISR) Instrument on NASA’s EOS Terra Satellite

MISR views of Eyjafjallajökull – 4/19/2010

Aerosol particle properties from MISR

7 7

MISR Version 22 Operational Aerosol Product for July 2007

Key Spherical Non-Absorbing

Spherical Absorbing Non-Spherical 8

Aerosol climatology from NAVY Aerosol Analysis and Prediction System (NAAPS) for July 2007

Note: AOT < 0.1 not shown 9

AERONET  AOD  550  nm  0   0.2   0.4   0.6   0.8   1  

0.2  

0.4  

0.6  

0.8  

1  

MISR  AO

D  550  nm

y  =  0.0275  +  1.061x      R  =  0.975  

MISR  Ocean  550  nm  

67%  within  expected  error  N  =  394  

0  

10

Salad

11 11

Motivation?

12 12

•  “Overall, about 70% to 75% of MISR AOD retrievals fall within 0.05 or 20% x AOD of the paired validation data from the Aerosol Robotic Network (AERONET), and about 50% to 55% are within 0.03 or 10% x AERONET AOD…” (Kahn et al., 2010)

Nature, 7 November 2013

Comparison  of  MISR  and  MODIS  

Particle/Mixture Issues

14 14

Kahn et al. (2009). MISR aerosol product attributes and statistical comparisons with MODIS, IEEE Transactions on Geoscience and Remote Sensing

GEOS-­‐5  Reanalysis  AcOviOes  

ICAP  6th  Working  Group  MeeOng,  Boulder,  CO,  October  21  -­‐  24,  2014  

Comparison  to  mulO-­‐year  satellite  data  sets  

Courtesy  of  Pete  Colarco  

Algorithm Updates (Completed)

19 19

Algorithm Updates (Completed)

20 20

•  Match angular resolution to SMART •  Update surface albedo threshold in AOD uppper bound to improve shallow

water coverage •  Replace quadratic interpolation with spline in SMART translation •  Enhance HET camera selection code to improve coverage in mountains •  Revised AOD upper bound mask to account for absorbing particles •  Require contiguous grid points in parabolic fit •  Calculate chi-square parameters at retrieved AOD (not on grid) •  Fix error in variance threshold •  Remove log transform in parabolic fit •  Lower floor in chi-square uncertainty •  Use glitter to retrieve windspeed over dark water (when possible) •  Add grid points at low AOD •  Eliminate AOD “uncertainty” as successful mixture criterion •  Increase Mie code iterations for Particle 6

Product Updates (Proposed)

21 21

Product Updates (Proposed)

22 22

• Move to 4.4 km x 4.4 km spatial resolution •  Separate surface algorithm from aerosol algorithm •  Simplify aerosol product

–  Separate product into (at least) USER and DIAGNOSTIC file

–  Include additional fields (e.g., lat/lon) to make data more user friendly

–  Critically examine contents of current product

MISR PIXLEYCA Local Mode Orbit 69644

Results – San Joaquin Valley, CA 20 Jan 2013

23 23

AERONET  DRAGON  Sites   24

N=14  

AERONET  DRAGON  Sites   25

N=14  

17.6  km  Standard  Product   26

4.4  km  Local  Mode  Product   27

4.4  km  Local  Mode  Product   28

MISR-­‐AERONET    nearest  Ome  to  MISR  overpass  

29

MISR  local  mode  4.4  km  resoluOon  

MISR  AO

D  (558  nm)  

AERONET  AOD  (558  nm)  

0                        0.05                0.10                  0.15                  0.20                  0.25                0.3                      

       0                      0.05                0.10                  0.15                0.20                  0.25                0.3                      

N=12  R=0.607845  R2=0.369476  RMSE=0.05461  

MISR-­‐AERONET    nearest  Ome  to  MISR  overpass  

MISR  local  mode  4.4  km  resoluOon  

MISR  AO

D  (558  nm)  

AERONET  AOD  (558  nm)  

0                        0.05                0.10                  0.15                  0.20                  0.25                0.3                      

       0                      0.05                0.10                  0.15                0.20                  0.25                0.3                      

N=13  R=0.918575  R2=0.843779  RMSE=0.02027  

30

O31948, P143 over India 20 Dec 2005

31

Comparison of 17.6 km and 4.4 km products

V22 17.6 km standard product

Prototype 4.4 km product

AOD

Comparison of 4.4 km prototype AODs with standard 17.6 km aerosol product

32

Main

33 33

0.001   0.010   0.100   1.000  

RelaOve  Num

ber  o

f  ParOcles  

MODIS  Small  ParOcles  

MISR  Small  ParOcles  ParOcle  Radius  (µm)  

0.001   0.010   0.100   1.000  ParOcle  Radius  (µm)  

RelaOve  Num

ber  o

f  ParOcles  

MODIS  Large  ParOcles  

MISR  Large  ParOcles  

RelaOve  Num

ber  o

f  ParOcles  

RelaOve  Num

ber  o

f  ParOcles  

0.01   0.10   1.000   10.00  ParOcle  Radius  (µm)  

0.01   0.10   1.000   10.00  ParOcle  Radius  (µm)  

Orbit  37813  Path  084  2007-­‐01-­‐26  

No  Successful  Matches  

Cameras  in  Glint  

Lines  Thick  =  ObservaOons  

Orbit  37813  Path  084  

2007-­‐01-­‐26  246m  –  Chisq  Abs  Floor  

2.0  m/s  Wind  NO  Geom  NO  Spec  BE  AOD:  0.0241  LR  AOD:  0.0224  

Passing  Mixtures:  246  LR  Mix:  69    

Spherical_Reff_0.12_Med_Dust_Coarse_Dust  

χ2abs  =  0.1370  

Lines  Thick  =  ObservaOons  

Thin  =  Retrieval  

Lines  Thick  =  ObservaOons  

Thin  =  Retrieval  

Orbit  37813  Path  084  2007-­‐01-­‐26  

246m  –  Chisq  Abs  No  Floor  2.0  m/s  Wind  NO  Geom  NO  Spec  

BE  AOD:  0.0220  LR  AOD:  0.0233  

Passing  Mixtures:  238  LR  Mix:  233  

Spherical_Reff_0.12_SSA_green_1.0_Reff_0.06_SSA_green_0.94_Absorbing  

χ2abs  =  0.2904  

MAN  =  0.0244  MISR  =  0.1496  

MODIS  =  0.0240  

MAN  =  0.0244  MISR  =  0.1496  

MODIS  =  0.0240  

MAN  =  0.0244  MISR  =  0.1496  

MODIS  =  0.0240  

MISR  Mixture  14  AErg  =  1.673  (1.673)  

AEmisr  =  1.621  

MAN  =  0.0244  MISR  =  0.1496  

MODIS  =  0.0240  

MODIS  ParOcle  1  AErg  =  2.726  (2.698)  

MAN  =  0.0244  MISR  =  0.1496  

MODIS  =  0.0240  

MISR  Mixture  14  AErg  =  1.673  (1.673)  

AEmisr  =  1.621  

MAN  =  0.0244  MISR  =  0.1496  

MODIS  =  0.0240  

MODIS  ParOcle  1  AErg  =  2.726  (2.698)  

74 Mixtures (Standard Product)

Key Spherical Non-Absorbing

Spherical Absorbing Non-Spherical

Spherical Absorbing + Non-Spherical (Tie)

Spherical Non-Absorbing + Non-Spherical (Tie) Spherical Absorbing + Spherical Non-Absorbing (Tie)

243 Mixtures (Ralph’s Set)

Key Spherical Non-Absorbing

Spherical Absorbing Non-Spherical

Spherical Absorbing + Non-Spherical (Tie)

Spherical Non-Absorbing + Non-Spherical (Tie) Spherical Absorbing + Spherical Non-Absorbing (Tie)

243 Mixtures (Ralph’s Set)

Key Spherical Non-Absorbing

Spherical Absorbing Non-Spherical

Smoke+Dust

Medium Mode Non-Absorbing Medium Mode Absorbing

Highly Absorbing

Standard V22 Het + Homog (246 Mixtures)

New Het (99%) + New Homog (Tau0 = 1.0)

Optimization

Chi-­‐Squared  Het  2-­‐D  Plots  

Dimensions:    x-­‐axis  =  AOD    y-­‐axis  =  Mixture  Number  (74  mixtures)  

1   2  Bad  Fit  0  Good  Fit  

AERONET  AOD  for  these  cases  is  shown  as  a  verOcal  green  line  Lowest  Chi-­‐Squared  Het  value  shown  as  a  symbol  (Asterisk)  Triplet  =  (Mixture,  AOD,  Chi-­‐Squared  Het)    Added  Upper  Bound  Mask  and  Lowest  Chi-­‐Squared  Value  (Triangle)  

Dessert

63 63

Benford’s  Law  •  Simon  Newcomb  (pictured)  published  an  arOcle  in  the  American  Journal  of  

MathemaOcs  in  1881  aeer  noOcing  that  books  of  logarithms  in  the  library  were  more  used  at  the  beginning    and  progressively  lesser  used.    He  inferred  that  scienOsts  were  looking  up  numbers  starOng  with  1  more  oeen  and  less  oeen  for  later  numbers.    He  expressed  this  mathemaOcally  with  the  equaOon:  

•  Frank  Benford  (not  pictured)  rediscovered  this  law  by  making  exactly  the  same  observaOon  about  books  of  logarithms  and  found  many  other  data  sets  that  followed  the  law.    He  published  his  results  in  a  paper  in  the  Proceedings  of  the  American  Philosophical  Society  in  1938.  

!

PD = log10 1+1 D( )

Benford’s  Law  

Sambridge  et  al.,  GRL,  2010  

Matched  MISR  V22/AERONET  Data  

Line shows predictions from Benford’s Law

Matched  MISR  V22/  

MODIS  Data  

References

68 68

Diner, D. J., et al. (2005), Using angular and spectral shape similarity constraints to improve MISR aerosol and surface retrievals over land, Remote Sens. Environ., 94, 155-171.

Kahn, R. A., et al. (2009), MISR aerosol product attributes and statistical comparisons with MODIS, IEEE Trans. Geosci. Remote Sens., 47, 4095-4114.

Kahn, R. A., et al. (2010), Multiangle Imaging SpectroRadiometer global aerosol product assessment by comparison with the Aerosol Robotic Network, J. Geophys. Res., 115, D23209, doi:10.1029/2010JD014601.

Diner, D. J., et al. (2012), An optimization approach for aerosol retrievals using simulated MISR radiances, Atmos. Res., 116, 1-14.

Kalashnikova, O. V., et al. (2013), MISR Dark Water aerosol retrievals: operational algorithm sensitivity to particle non-sphericity, Atmos. Meas. Tech., 6, 2131-2154.

Sambridge, M., et al. (2010), Benford’s law in the natural sciences,

Geophys. Res. Lett., 37, L22301, doi:10.1029/2010GL044830.